The unending obesity panic in the face of the data

We haven’t had a thread about this in quite awhile, and I realized it’s been twenty-five years now since I started arguing with people like S. Jay Olshansky and David Ludwig about the “obesity epidemic.” At the time, I would point out that there was simply no evidence that their predictions about the dire effects of the explosion of fat people on mortality rates were coming true. The response was always some variation on “just you wait and see, it may be true there’s no evidence now (also I’m a doctor and you’re a lawyer so your arguments are not to be taken seriously by definition), but there will be such evidence, eventually if not sooner.”

Well it’s a quarter century later, and . . . still no evidence! I mean it’s truly remarkable: We now have 65 years of good data on the changing rate of “obesity” in the US population, and considerably longer-term data than that on the changing rate of the age-adjusted mortality rate of that same population, and it couldn’t be clearer that there is absolutely no correlation between the two variables whatsoever. In other words, the rate of decline in age-adjusted mortality doesn’t show even any correlation with changing obesity rates, let alone any causal relationship. That rate of decline stays the same when obesity rates are skyrocketing, compared to when they’re remaining flat.

Quick definition: Age-adjusted mortality rates for the population as a whole are always measured against what’s known as a standard population, which is to say an idealized population that maintains the same age ratio over time. If you didn’t do this, then mortality rates would appear to be much lower in Niger (median age 15) than in Japan (median age 50 — that’s right, 50).

What I’ll call the obesity mortality hypothesis is that if you have a country in which the obesity rate stays more or less flat for a quarter century, and then skyrockets in the next quarter century, then you should observe either:

Obesity mortality hypothesis, strong version: Sharply declining age adjusted mortality rates in the first twenty five years, and sharply increasing age adjusted mortality rates in the second twenty five years; or, in the alternative

Obesity mortality hypothesis, weak version: A significant slowing in the rate of decline in age adjusted mortality over the the twenty five years when obesity rates are skyrocketing, relative to the twenty five years when they were flat.

Notice that the weak version of this hypothesis is pretty damn weak: It acknowledges that obesity can’t be that big of a variable in mortality risk, because it admits that even skyrocketing obesity rates at a population wide level will merely lead to some slowing in mortality rates, rather than a flattening of them, let alone an actual reversal, i.e., increasing age adjusted mortality rates (this is the strong hypothesis, which is what Ludwig, Olshansky et al were arguing was going to happen twenty-five years ago).

Let’s go to the stats. By far the best data we have on obesity rates is the NHANES survey, which has been conducted intermittently since 1960. That data show that obesity rates were basically flat in the 1960s and 1970s (13.4% in 1960; 15% in 1980) before beginning to climb rapidly. Unfortunately the next NHANES survey wasn’t complete until 1994, but if we look at the change in age adjusted mortality between 1960 and 1985, when statistically speaking there simply was no “obesity epidemic,” we find that this fell by 26.2% from 1339/100K to 988/100K. So what happened in the next quarter century, when the “obesity epidemic” exploded? Obesity rates went from 15% in 1980 (we don’t have 1985 data), to 36.1% in 2010 — a startling 141% increase in the prevalence of this deadly condition. According to the strong hypothesis we should have seen a massive spike in age adjusted mortality over this time frame, while the weak hypothesis would anticipate a rapid slowing in the decline in age adjusted mortality.

Instead, what we see is a decline in age adjusted mortality from 988/100K to 747/100K — in other words, practically the identical percentage decline to that during the golden years of no obesity epidemic anywhere in sight (24.4$ v. 26.2%).

In fact, if you take a look at the last 75 years in terms of quarter century increments, the steadiness in the decline in the age adjusted mortality rate in the US is remarkable: 24% from 1950 to 1975, 21% from 1975 to 2000, and 21% from 2000 to 2025, despite a big spike during the pandemic years. (This of course also illustrates the complete absurdity of the entire MAHA discourse: you were on average less than half as likely to die over the next year in 2025 than you would have been at the same age in 1950. Age adjusted mortality rates in the two years: 1,446/100K; 689/100K).

In other words, even the weak version of the obesity mortality hypothesis remains completely unsupported by any population-wide data: the decline in mortality rates has been identical in national populations in which the obesity rate more than doubled, compared to national populations in which it remained essentially unchanged over the same length of time.

But hey, maybe the next quarter century will start to produce the very first evidence for the weak hypothesis, if not the strong one. On the other hand J.M. Keynes had something to say about the long run. (They don’t call it the dismal science for nothing).